PrimeIntellect/INTELLECT-2
TEXT GENERATIONConcurrency Cost:2Model Size:32.8BQuant:FP8Ctx Length:32kPublished:May 11, 2025License:apache-2.0Architecture:Transformer0.2K Open Weights Cold

INTELLECT-2 is a 32 billion parameter language model developed by PrimeIntellect, trained using a distributed asynchronous reinforcement learning framework. Based on the Qwen2 architecture, it specializes in mathematical and coding tasks, demonstrating improved performance over its base model, QwQ-32B, in these domains. The model was trained leveraging community-contributed GPU resources and features a 131072 token context length.

Loading preview...